A DEFENSE OF THE SEMANTIC VIEW OF THICK CONCEPTS
Miriam Johnson
A thesis submitted to the faculty at the University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree of Master of Arts in the Philosophy department in
the College of Arts and Sciences.
Chapel Hill 2020
Approved by:
Sarah Stroud
Geoff Sayre-McCord
ABSTRACT
Miriam Johnson: A Defense of the Semantic View of Thick Concepts (Under the direction of Sarah Stroud)
Pekka Vayrynen (2013) argues that the significance often attributed to thick ethical terms
in philosophy depends on a false assumption: that the evaluations conveyed by our use of these
terms is part of their semantic content. Contrary to this “semantic view,” Vayrynen claims, thick
evaluations are best thought of as a species of pragmatic implication, meaning that evaluation is
not an essential feature of most thick concepts. This paper responds to Vayrynen’s core thesis
and argues that it ought to be rejected. Contrary to Vayrynen’s claims, the pragmatic view does
not more accurately capture the linguistic data that he cites, nor does it offer a simpler
explanation of thick evaluation than the semantic view. This paper does not offer a substantive
defense of the metaphysical significance of thick terms, aiming instead to defend thick terms as a
ACKNOWLEDGMENTS
Thanks so much to Sarah Stroud, Geoff Sayre-McCord, and Gillian Russell for their
TABLE OF CONTENTS
CHAPTER 1:INTRODUCTION ... 1
CHAPTER 2:THE PRAGMATIST’S CHALLENGE ... 5
SECTION 2.1 THE APPROPRIATENESS CLAIM ... 6
SECTION 2.2 THE SIMPLICITY CLAIM... 8
CHAPTER 3:AGAINST THE APPROPRIATENESS CLAIM ...10
SECTION 3.1 THE NORM-RELATIVE SEMANTIC VIEW ...10
SECTION 3.2 WHAT ARE NORM-RELATIVE THICK CONCEPTS? ...11
SECTION 3.3 CRITIQUING THICK JUDGMENTS AND EXPRESSIONS ...14
SECTION 3.4 EXPLAINING THE PRAGMATIST’S DEFEASIBILITY DATA ...17
SECTION 3.5 NORM RELATIVE TERMS VS. DESCRIPTIVE TERMS...20
CHAPTER 4:AGAINST THE SIMPLICITY CLAIM ...24
SECTION 4.1 THE NORM-RELATIVE VIEW AND THE IS-OUGHT GAP ...24
SECTION 4.2 IMPLICATIONS...26
SECTION 4.3 THE ROBUST NORMATIVITY OBJECTION ...28
SECTION 4.4 THE OBJECTION FROM CORRECTNESS CONDITIONS ...29
SECTION 4.5 CONVENTIONAL ...30
SECTION 4.6 AGAINST CONVENTIONAL ...32
SECTION 4.7 IMPLICATIONS OF REJECTING CONVENTIONAL ...34
CHAPTER 5:CONCLUSION ...35
CHAPTER 1: INTRODUCTION
A number of prominent philosophers have argued that if arguments like the following are
valid, this has important metaethical implications.1
(1)
(a) x causes offence by indicating a lack of respect. (b) x is rude. 2
In particular, (1) undermines the widely held assumption that a descriptive premise cannot
on its own (without the addition of at least one evaluative premise) entail an evaluative
statement--or rather, that we can’t derive an “ought” from an “is.” What is puzzling about (1), then, is that
we can point to no logical problem with it, and yet, (b) seems to be a value judgment, while (a) is
clearly not. (1) appears to violate the rule outlawing derivations from “ought” to “is” in the sense
that we can infer from (b) that
(c) x is pro-tanto bad in some way.
If this is correct, it is a problem in particular for metaethicists who claim that evaluative
terms are semantically distinct, having some feature f (involve emotions, attitudes, the acceptance
of imperatives) that distinguishes them from descriptive terms (call this assumption Semantically
1 See Foot (1958), Searle (1969), and Anscombe (1958).
Distinct), This is because if evaluative language is distinct in this way, “no word or statement
which does not have the property f can be taken as equivalent to any evaluation, and no account
of the use of an evaluative term can leave out f and yet be complete.”3 More succinctly, if
evaluative language is distinct from descriptive language in this way, then (a) and (b) cannot be
equivalent if only (b) is evaluative. Metaethicists who deny that we can infer an “ought” from an
“is,” then, usually resist the inference from (a) to (c) by denying that (b) is evaluative. According
to the non-cognitivist, for instance, (b) is an ordinary descriptive term that we sometimes use to
express evaluations. Thus, (b) is only an evaluative premise if we conjoin the term “rude” with
some non-cognitive attitude.
There is a debate between people who think that RUDE is inherently evaluative and thereby
that (1) undermines Semantically Distinct, and people who resist this conclusion by denying that
RUDE is inherently evaluative. Call the former view the semantic view, because it assumes that
thick concepts like RUDE are evaluative in the sense that they semantically entail (as opposed to
pragmatically implicate) evaluations. Call the latter the pragmatic view, since it says that even
though thick concepts like RUDE only appear evaluative when they really aren’t. According to
the pragmatic view, the evaluations conveyed when we use thick terms is not due to the meaning
of these terms, but rather to preexisting evaluative assumptions held by speakers in a given
linguistic community.
There are a number of reasons to think that the pragmatic view of thick terms and concepts
is the better option. Not only does it explain our intuition that terms like “rude” are evaluative, but
it does so without committing us to the idea that (1) undermines the is/ought gap. Further, the
pragmatic view appears to be able to better explain the way that we use thick concepts, as compared
to the semantic view. In particular, it explains—where the semantic view apparently cannot--our
sense that the evaluative content of thick terms is defeasible.
The main goal of this paper will be to argue that these apparent advantages are deceptive
and that the semantic view is ultimately preferable to the pragmatic one. I’ll do this by way of
defending a version of the semantic view of thick concepts that is rarely discussed--I’ll call it the
norm-relative semantic view. The norm-relative semantic view, I’ll argue, not only does a better
job explaining the defeasibility data alluded to above, but it also offers a nice way to explain what
is going wrong in inferences like (1) without denying that thick terms and concepts are inherently
evaluative. Finally, the norm-relative semantic view explains better than the pragmatist can,
exactly how descriptive and normative concepts differ with respect to their semantics.
The structure of the paper will go as follows: First, I’ll focus on defending the norm-relative
semantic view against the pragmatic view of thick concepts. I’ll start with a summary of why the
pragmatic view claims to be preferable to the “default semantic view.” Then I’ll introduce the
norm-relative semantic view, and show how it avoids the pragmatist’s objections to the default
semantic view. Finally, I’ll argue that not only is the norm-relative semantic view able to
accommodate the supposedly problematic data, but it is also preferable to the pragmatic view along
the lines of simplicity and explanatory power.
The final two sections of the paper will explain how the norm-relative semantic view
bears on the larger issues at play in the debate over thick concepts. I’ll argue here that the
is, but that, at the same time, she can also explain how thick concepts can be both norm-relative in
the sense described and subject to higher order correctness conditions. If I am correct, then thick
concepts are not as important for telling us about the nature of value as many people have claimed,
but neither should they be treated as insignificant by value theorists. Recognizing thick concepts
when we find them is important most of all, for helping us better understand our own evaluative
CHAPTER 2: THE PRAGMATIST’S CHALLENGE
As mentioned, it is standardly assumed that whether (1) undermines the fact-value
distinction is contingent on whether “rude” is inherently evaluative. Because “rude” is a term that
seems to at once be both descriptive and evaluative—a “thick” term/concept—it isn’t obvious
whether it really is both descriptive and evaluative or whether, as the pragmatist argues, it is just a
descriptive term that we often use to make evaluations. Therefore, it is likewise not obvious
whether (1) really is an inference from “ought” to “is.”
For the purposes of this paper, we can assume that the semantic view of thick concepts is
the default. This is the default view because we can agree that “rude” is a term that we use to
evaluate; it justifies a certain kind of condemnation that we might have for certain behavior. It
therefore seems only reasonable to think that “rude” is, despite its obvious descriptive content, an
evaluative term. According to the semantic view, then, the meaning of “rude” is something like
“being pro-tanto bad in virtue of causing offense by indicating lack of respect.”
The pragmatic view, then, is the skeptical response to this intuition. Straightforwardly, the
pragmatic view says that there is an alternative—and better—explanation available for why
arguments like (1) seem evaluative. “X is rude” indicates that X is bad in a certain way by way of
the same mechanism that “X is a bird” indicates that X can fly, or “X is a cat” indicates that X has
fur; namely, it is pragmatically implicated. In the same way that being able to fly or having fur are
concept “rude.” Being a bad thing, then, is conversationally implicated when we say things like
“X is rude.”
The pragmatic view appears to be superior to the default semantic view in at least two
ways. First, it the pragmatic view claims to be better situated than the default semantic view to
explain the behavior of evaluations associated with thick terms in certain kinds of utterances. In
particular, the Pragmatist claims that the evaluative information conveyed by expressions using
thick terms is cancelable in a way that the semantic view cannot as easily explain. Second, the
pragmatic view is simpler. It gives us an easy way to explain (1) without needing to draw any
metaphysical conclusions. According to the pragmatic view, “rude” is a descriptive concept, so
there is no illicit inference from “is” to “ought.” Second,
Section 2.1 The appropriateness claim
According to the pragmatist’s setup, the semantic view says that thick concepts are
inherently evaluative in the sense that “rude” means something like “being bad in virtue of having
descriptive features D.” The semantic view, then, says that the sentence
(2) S is rude, but not bad in any way.
is equivalent to
(3) S is bad in virtue of having descriptive features D, but S isn’t bad in any way.
meaning that we ought to read it as a contradiction. This is because, on the semantic view,
evaluation is part of a thick term’s truth-conditional content. But, the pragmatist points out,
We can test whether the evaluative information conveyed by thick terms is part of their
semantic content or conveyed by way of conversational implicature, the pragmatist argues, by
asking whether the evaluation can be canceled. In other words, if X is semantically entailed by a
sentence, then we shouldn’t be able to cancel X by following the sentence with the disclaimer,
“but I didn’t mean to imply that X.” If the semantic view is correct, then the following should be
contradictory (in the same way that “S has arthritis, but I don’t mean to imply that she has a joint
condition” or “X is illegal, but I don’t mean to imply that it’s against the law” are contradictory):
(5) Whether or not Madonna's show is lewd, it's not bad in any way.
(6) Whether or not Madonna's show involves explicit sexual display, it would be in no
way bad for that.4
(7) Whether or not this is good in any distinctive way, John can be truthfully and
neutrally described as being considerate.5
Because the semanticist defines LEWD as “being sexually provocative and therefore bad in
a certain way,” (5) and (6) should, if the semantic view is correct, strike us as contradictory. The
same goes for CONSIDERATE in (7), which, on the semantic view, entails that John is good in a
certain way.
The pragmatic view, of course, can explain why (5)-(7) don’t strike us as contradictory,
since, on her view, there is nothing contradictory or infelicitous about them. If evaluative content
is pragmatically implicated, then it is, by definition, cancelable. In other words, for any pragmatic
4 Ibid. p. 70
implicature such as the implication of “S has very good handwriting,” in a recommendation letter,
that John is unimpressive, can be canceled. S’s letter writer could, without any linguistic error or
infelicity, qualify his mention of S’s handwriting with something like “but I don’t mean to imply
that S is unimpressive.” This, the pragmatist thinks, is what is going on in (5)-(7). In each case,
the thick term implicates an evaluative judgment about Madonna’s concert or about John, but the
second clause cancels that implication, and this strikes us as not unnatural.
For the pragmatic view, then, saying that Madonna’s show is lewd is not semantically distinct
from saying that it involves explicit sexual display. This means that (5), for the pragmatist, is
equivalent to
(8) Whether or not Madonna's show involves explicit sexual display, it's not bad in any way for that.
Indeed, we might even think that the second clause cancels any implication that the show’s
involving explicit sexual display is somehow bad.
Section 2.2 The simplicity claim
Call the pragmatist’s claim that the pragmatic view offers the best explanation of the
defeasibility data discussed above the appropriateness claim, since it says that, at least when it
comes to utterances like the ones mentioned above, the pragmatic view offers a better explanation
for what is going on.
But before I go on to respond on behalf of the semantic view, it is important to highlight a
background assumption held by the pragmatist, which will prove important if we want to defend
the semantic view against the pragmatist’s challenge. This is the assumption that the pragmatic
due to the fact that, unlike the pragmatic view of thick evaluation, the semantic view has heavy
duty metaphysical implications. There is a background assumption held by metaethicists, in other
words, that we have independent reason—aside from the pragmatist’s linguistic data—to reject the
semantic view—namely, that it is metaphysically problematic. Call the idea that the pragmatic
view is preferable to the semantic view because the former lacks the latter’s metaphysical baggage
the simplicity claim.
In order to fully disarm the pragmatist’s challenge to the semantic view, then, the defender
of the semantic view needs to show both that she can explain the linguistic data just as well as the
pragmatist can, and that she can do this without accruing any significant metaphysical
commitments. Thus, she needs to show that the pragmatic view is not more explanatory than the
semantic view and that the pragmatic view is not simpler than the semantic view. In the following
chapter, I’ll argue that the pragmatic view is not more explanatory; in the fourth chapter, I’ll argue
CHAPTER 3: AGAINST THE APPROPRIATENESS CLAIM
Thus far we have two views: the default semantic view, which says that thick terms and
concepts are semantically evaluative and that (1) is therefore an inference from “ought” to “is,”
and the pragmatic view, which says that thick terms and concepts are merely descriptive, making
(1) unimportant. If the above is correct, the pragmatist seems to have the better view. In this
section, however, I’ll argue that the pragmatist’s attack on the semantic view portrays it in a way
that is too strong. A slightly weaker version of the default semantic view can not only handle the
pragmatist’s objections, but it is more explanatory of the data on thick terms and concepts.
Section 3.1 The norm-relative semantic view
As mentioned, the semantic view is defined by the pragmatist according to its
commitment to the semantic equivalence of (3) and (4).
(3) S is rude, but not bad in any way.
(4) S is bad in virtue of having descriptive features D, but S isn’t bad in any way.
But this definition of the semantic view is importantly ambiguous, for the following reason.
We can interpret “bad” in (3) and (4) in different ways. First, we could interpret “bad” in a way
that is relativized to some specific normative standard such as, for instance, the standards included
in conventional rules of etiquette. This means that we could read (3) as
But we could also read “bad” as unsubscripted. If we do this, then it will be natural to read
the second clause of (3) as making an all-things-considered or robust judgment about S, given that
she is rude. This is importantly different from making a narrow evaluative judgment about S in the
domain of etiquette. We would then read (3) as
(10) S is rude, but not bad in the unsubscripted, all-things-considered sense.
Therefore, we can make a prima facie distinction between two versions of the semantic
view. First, there is the default semantic view which treats the “bad” in (3) as unsubscripted.
Second, there is the norm-relative semantic view, which treats “bad” in (3) as subscripted, or
relativized to some specific normative standard or set of standards such as the local rules of
etiquette, or the rules of chess or soccer.
Importantly, (9) and (10) give us different intuitive results regarding the incoherence
question. If we interpret the semantic view in the default way, then we’ll read (3) as (10), which
does not seem contradictory. If we interpret the semantic view in the norm-relative way, we’ll read
(3) as (9), and (9) does seem incoherent since what it means for someone to be rude is for her to
violate a norm of politeness; even if my rude behavior isn’t necessarily bad in any robust (e.g.,
moral) sense, it doesn’t follow that it wasn’t bad with respect to local rules of etiquette.
From this much, it should be clear that there are two ways to interpret the semantic view.
Now it needs to be made clearer why we should interpret thick concepts in the norm-relative sense.
Section 3.2 What are norm-relative thick concepts?
According to the norm-relative semantic view, a concept is semantically evaluative if
restricted to absolute moral or epistemic standards such as “the Good” or “the True.” Instead, they
include legal standards, standards of etiquette, standards for good and bad moves in a game,
standards for behavior within religious communities, standards for good or bad apples, and so on.6
Given the permissiveness with which this view understands what it means to be an evaluative
standard, it might seem unclear how exactly we can differentiate semantically thick from merely
pragmatically evaluative terms. To get a clearer sense of the intuitive distinction, consider the
difference between
(11) WALKING ACROSS THE STREET WITHOUT A CROSSWALK and
(12) JAYWALKING, and between
(13) PAYING X DOLLARS IN TAXES, vs.
(14) Engaging in TAX EVASION or
(15) TAX AVOIDANCE.
(11) and (12) have the same non-evaluative content: namely, they describe a way that a
pedestrian might cross a busy road. Only the latter, however, evaluates the relevant action with
respect to an evaluative standard. In this case, it is the standard of traffic laws. Whereas the content
of (11) is nothing other than a description of a certain action, (12) refers to that action while also
evaluating it with respect to the traffic law, specifically, specifying that this action violates traffic
laws and is therefore “bad” with respect to those laws. We can think of the (13) vs. (14) and (15)
6 The normative status of these latter standards is what some philosophers might call “formal” or
in a similar way. While the former describes an action without evaluating it, the latter two concepts
describe that action while specifying whether or not it is legally permissible. 7
(12), (14), and (15) are examples of concepts whose content contains evaluation with
respect to a specific system of rules. But we can also find concepts in our vocabulary that contain
evaluation with respect to a specific standard of taste. Consider, for instance, the terms and
concepts employed by wine tasters to describe different taste features of a given wine. Adrienne
Lehrer’s Wine and Conversation makes the following observations about the language used by
professional wine tasters. First, she observes that “wine words can be analyzed in terms of various
dimensions: acidity, sweetness, body, balance, feel, age, nose, finish, activity, and quality.”8 But
along each dimension, she finds, there are terms referring to flavors that constitute “too much”
“too little” and “right amount” of the relevant property. For instance, along the dimension of
acidity, there are a collection of terms used to indicate “too much acidity.” These include hard,
BITING, and SHARP. Tastes indicating too little acidity include FLAT and BLAND. Tastes
indicating the right amount of acidity include TART, CRISP, and TANGY.9
It is important here that these concepts would not refer to the properties they do if not
for the wine tasters’ determination that “taste X constitutes the right amount of acidity.” That is,
if evaluation did not factor into the process, the closest alternative to this particular scale would
7 If there is a case to be made that “green” serves to rank a certain color property according to an
evaluative standard of some kind, then perhaps “green” would be a thick concept. But for now, I’m going to assume that there is an obvious enough distinction between terms that are formally evaluative and those that aren’t.
8 Lehrer 2009 p.7
include “more,” “less,” and “in-between” degrees of acidity. But, plausibly, this range of terms
would not be descriptively identical to the terms just listed, which constitute flavors identified on
the basis of evaluation.
What is distinct about these concepts is that they constitute a method for transmitting
value judgments about wine. Thus, if I, someone who does not have this level of sophistication,
am taught this wine vocabulary, I will, in virtue of learning these terms, also have adopted a set of
evaluative beliefs. That is, I will, once I’ve learned this vocabulary, be able to tell someone that
tart, mellow, and crisp wine is good; that rough, course, thin, or unripe wine is bad.
Section 3.3 Critiquing thick judgments and expressions
An important feature of the norm-relative view is that it explains why we might
frequently engage in “higher order” disagreements about our judgments using thick concepts. In
Vayrynen’s (2013) defense of the pragmatic view, he addresses the question of whether the
defeasibility data in 2.1 might be able to be explained as a certain kind of higher order disagreement
about the meaning of concepts like “lewd.” If this was the case, then we would interpret the second
clause of sentences like (5) as critiques of the concept “lewd;” as the claim that “lewd” should
mean something other than it does.10 This, Vayrynen claims, is implausible, for a few reasons:
[First] It isn't clear how this explanation is supposed to capture lewd-objectors' lack of interest in disputes about how explicit a sexual display must be to count as satisfying lewd. [Second] even if the hypothesis gave a good explanation of the defeasibility data, it wouldn't explain the projection data. Suppose that lewd-users agree that the meaning of lewd doesn't contain evaluations… but go on using lewd anyway. It would be reasonable
to interpret exchanges among lewd-users as still conveying evaluations.11
The norm-relative semantic view, however, can give a plausible explanation for how these
disagreements work that avoids Vayrynen’s objections. According to the norm-relative semantic
view, disagreements about the meaning of “lewd” will not be about whether “lewd” is an
evaluative term, but rather will be about whether and how the norm that “lewd” expresses is
justified.
To get a clearer sense of the way that higher order disagreements work on the norm-relative
semantic view, consider a concept like FDA-APPROVED. On the norm-relative view, there are at
least two ways for me to critique the sentence
(16) Drug x is FDA-approved.
At the first order, I can ask whether drug x meets the current criteria provided by the FDA.
At the second order, however, I can ask whether FDA-approved is a reliable standard for judging
the safety of drug x. Suppose, for instance, that I have a conflicting source which says that drug x
is actually dangerous. In this case, I might discover that the FDA is actually wrong about the safety
of drug X, and therefore I might campaign to get the FDA to revise its guidelines in light of the
evidence that I’ve discovered.
This second-order critique of (16) is quite different from the one that Vayrynen gestures at
above. According to Vayrynen, if the second clause of (5) is a higher order critique of the concept
“lewd,” then it must be contesting whether “lewd” is inherently evaluative. But according to the
norm-relative view, the critique of “lewd” in (5) is analogous to the critique of FDA-approved just
described. That is, we can read the second clause of (5) as a challenge to the assumption that the
evaluative standard according to which things are lewd is justified. We might say, for instance,
that a strictly utilitarian approach to etiquette norms where “lewd” indicates an inappropriate
display of sexuality would give us different results regarding what counts as “lewd,” and this would
likely result in our concluding that Madonna’s show was not lewd.
We might also be criticizing the idea that lewdness is a legitimate normative standard, in
the same way that I might say that
(17) X is FDA approved, but the FDA is an unreliable source of information regarding x.
I might also say that
(18) X is lewd, but X ought not to be considered lewd, since X isn’t really bad.
According to the semantic norm-relative view, what we mean by (18) is that
(19) X is rude according to conventional standards of propriety, but X ought not to be considered rude, since the conventional standards of propriety need revision.
It is worth emphasizing that we seem to often engage in this kind of higher order critique
of thick concepts when we want to express dissatisfaction with rules or norms that we think are
unreasonable. We often say things like
(20) X is beautiful by conventional standards, but he isn’t really beautiful.
(21) X is a traffic violation, but it isn’t really wrong.
Section 3.4 Explaining the pragmatist’s defeasibility data
Recall that the pragmatist’s main argument against the semantic view says that she can’t
make sense of the coherence of (5)-(7). According to Vayrynen, what is going on in these sentences
is that in each case the speaker is canceling their evaluative content in the second clause. This
objection might work against the default semantic view, but it doesn’t work as an objection against
the norm-relative semantic view. Start with the pragmatist’s main cases:
(5) Whether or not Madonna's show is lewd, it's not bad in any way.
(6) Whether or not Madonna's show involves explicit sexual display, it would be in no
way bad for that.12
(7) Whether or not this is good in any distinctive way, John can be truthfully and
neutrally described as being considerate.13
As a first step, recall the argument in section 3.3 that if the norm-relative view is correct,
then there are two ways to critique (5)-(7). Thus, we need to start by assuming the norm-relative
view, which means that by “lewd,” we mean something like “sexually explicit and therefore bad
according to the conventional standard of propriety.” We would therefore interpret the first clause
of (5) as follows:
(23) Whether or not Madonna's show is [sexually explicit and therefore bad according to conventional standards of propriety] it's not bad in any way.
12 Ibid. p. 70
Here’ we’ve assumed the norm-relative view of “lewd,” meaning that its semantic content
contains evaluative information with respect to a standard. This can be distinguished from the
default semantic view, which would treat (5) as equivalent to
(24) Whether or not Madonna's show is [sexually explicit and therefore bad [unsubscripted] it's not bad in any way.
(24), interpreted along the lines of the default semantic view, is incoherent. However,
(23) is not, on the condition that we read “bad” as unsubscripted.
(25) Whether or not Madonna's show is [sexually explicit and therefore bad according to conventional standards of propriety] it's not bad [unsubscripted] in any way.
According to the norm-relative semantic view of thick concepts, since these concepts refer
to normative standards that are open to critique, it makes sense to think that we naturally read (5)
as equivalent to (25). This is because the second clause of (25) seems clearly to constitute a critique
on the part of the speaker of the relevant conventional standard of propriety. The speaker is, in
other words, acknowledging that Madonna’s show was lewd according to conventional standards,
but pointing out that it was not thereby bad in any robust sense.
The goal of this subsection has been to show not that this is definitively the correct way to
read (5), but rather that there is a reasonable way for the norm-relative semantic view to explain
why (5)-(7) are not incoherent. Specifically, she can argue that what is going on in (5)-(7) is that
the speaker is giving a second order critique of the standards to which each thick term appeals
(e.g., the conventional standard of propriety, or the conventional standard of virtue).
This means that we now have two competing interpretations of (5)-(7): the pragmatist’s
explanation, which says that the second clause of (5) cancels the implication that lewdness is
criticizes the standard of propriety according to which Madonna’s show counts as bad (in this
Section 3.5 Norm relative terms vs. descriptive terms
Suppose that the two explanations of (5)-(7)—the pragmatist’s explanation and the
norm-relative semantic view’s explanation—are equally plausible. Even if this is the case, I think that
we have clear reason to prefer the norm-relative semantic view to the pragmatic view. This is that
there is, contrary to what the pragmatist argues, a clear difference between ordinary descriptive
terms that we often use to evaluate people like “extraverted” or “tall” and terms that are necessarily
evaluative, such as “courageous” or “chaste” or “rude.” A good way to distinguish the latter terms
from the former is to say that the latter presuppose an evaluative standard—a standard according
to which we rank things as good or bad, better or worse, allowed or impermissible.
According to the pragmatic view, there is no important difference between the way that
evaluation is conveyed by the term “extraverted” as compared to the term “courageous.” But what
does the pragmatist mean here by “evaluation”? As discussed above, there are at least two ways
of understanding “good” and “bad,” and, as a result, there are different ways of understanding
“evaluation.” This becomes clearer when we consider a term like “jaywalk.” As mentioned earlier,
the term “jaywalk” is defined as something like “to cross or walk in the street or road unlawfully
or without regard for approaching traffic.” How does the pragmatist treat a term like “jaywalk”?
The norm-relative semantic view has an obvious account of what this term means: it is, on
the norm-relative semantic view, a thick term, in the sense that its semantic content evaluates
something with respect to a normative standard. In this case, that standard is that imposed by
certain traffic laws. But is “jaywalk” a thick term on the pragmatic view? Consider the claim that
Given that this sentence is structurally analogous to (5)-(7), and that it seems coherent and
natural, it makes sense that the pragmatist would interpret the second clause of (26) as canceling
the negative evaluation implicated by “jaywalk.” But if this is correct, then there shouldn’t be a
semantic difference between (26) and
(27) X counts as crossing the street without regard for traffic, but it isn’t bad in any way.
Assuming that we generally think it is bad to cross the street without regard for traffic, (26)
and (27) will be, for the pragmatist, equivalent. But this cannot be right, because (26) includes
important information that is not included in (27), namely, that X is an illegal act. That is, suppose
I tell you either that
(28) S crossed the street without paying attention to oncoming traffic on her way here.
Or
(29) S jaywalked on her way here.
We can infer from the semantic content of (29) that S violated traffic laws, but we cannot
infer this from (28), unless we presuppose that there are traffic laws outlawing this kind of careless
pedestrian behavior. To assume that “jaywalk” does not include information with respect to
violating traffic laws is to assume that “jaywalk” and “cross the street without paying attention to
oncoming traffic” mean the same thing. But this can’t be right. Here is one explanation for why.
Some of our language depends for its meaning on sets of norms or rules. John Rawls’
(1955), for instance, gives the example of baseball terms. According to Rawls,
walk, or make an error, or balk; although one can do certain things which appear to resemble these actions such as sliding into a bag, missing a grounder and so on. Striking out, stealing a base, balking, etc., are all actions which can only happen in a game. No matter what a person did, what he did would not be described as stealing a base or striking out or drawing a walk unless he could also be described as playing baseball, and for him to be doing this presupposes the rule-like practice which constitutes the game. The practice is logically prior to particular cases: unless there is the practice the terms referring to actions specified by it lack a sense. (25, italics mine)
We can say something analogous, to what Rawls says here about baseball terms about the
term “jaywalk.” In the same way that we can coherently talk about “throwing a ball,” “running,”
or “swinging a peculiarly shaped piece of wood” without presupposing the game of baseball, we
can also coherently talk about “crossing the street without regard for traffic” without presupposing
any evaluative standard. By contrast, however, to talk about “stealing a base,” or about
“jaywalking,” one must presuppose the game of baseball and traffic laws respectively. We can, in
other words, say that traffic laws are logically prior to the concept JAYWALK.
The idea that many thick concepts are dependent in some necessary way on preexisting
normative standards is even clearer when we consider a variety of other thick concepts like
JAYWALKING, TAX EVASION, FDA-APPROVED, SHARP and BITING. In each case, there
is a clear difference between the thick term and what, on the pragmatic view, would be its
descriptive equivalent. When we say that a wine is “biting,” we mean that it is too acidic, not that
its acidity is high. As mentioned earlier, these concepts plausibly refer to different properties, since
what constitutes high acidity will not correspond to it if tasters prefer wine that is highly acidic,
and use “biting” to refer only to wine with extremely high acidity for instance.
A helpful way to think about the difference between inherently evaluative concepts
and descriptive concepts that we use to evaluate is by comparing them to tools. Consider two
The rock is not essentially a tool used to pound nails. Instead, it’s an object that I’ve made use of
in this particular way, but whose essential function isn’t to pound nails. The hammer, by contrast,
does have the essential function of pounding nails. Plausibly, this difference, between the object
that is merely used to pound nails and the object that is designed to pound nails is the same kind
of difference that we see between the concept JAYWALKING and the concept CROSSING THE
STREET WITHOUT REGARD FOR TRAFFIC, and the concept of being HIGH IN ACIDITY,
CHAPTER 4: AGAINST THE SIMPLICITY CLAIM
Suppose that, at least on the basis of linguistic data, we have good reason to prefer the
norm-relative semantic view to the pragmatic view of thick concepts. Even if this is the case, there
is still the problem of thick concepts and the is/ought gap discussed in the introduction.
Section 4.1 The norm-relative view and the is-ought gap
Recall that the pragmatic view, unlike the default semantic view, gives us a way to explain
what is wrong with arguments like (1) (here it is again):
(1)
(a) x causes offence by indicating a lack of respect. (b) x is rude.
Recall that philosophers have mostly assumed that if (1) does not undermine the is/ought
distinction, then it is because (b) is not evaluative, since “rude” is not inherently evaluative. What
this means is that, prima facie, if the norm-relative semantic view is correct, then we’ll have to
accept that (1) does constitute a valid inference from is to ought, and this will be a problem for the
many people who deny that this is possible.
But the norm-relative semantic view gives us an alternative way to explain what is
wrong with (1), which doesn’t require the pragmatic view. Recall that the norm-relative semantic
view says that thick concepts get their evaluative content on the basis of some preexisting
standard of politeness according to which x counts as rude. Compare (1) to a different, but similarly
structured, argument:
(30)
(a) x is an orange striped ball that we use to throw into a netted hoop. (b) x is a basketball.
Does (30) give us conclusive reason to think that basketballs exist in a robust sense, in the
same way that trees and rocks exist? No, because (30b) only follows from (30a) if we assume that
there is a game of basketball. Similarly, (1b) only follows from (1a) if we assume that there is an
existing norm according to which lack of respect is considered bad. What this means is that even
if we can infer from (1a) and (1b) that
(30) x is bad in a certain way.
We still won’t have undermined the is/ought gap. This is because, on the norm-relative
semantic view, (1b) only gives us permission to infer that
(31) x is bad with respect to conventional (or some other) norms of politeness.
In other words, we don’t get any sort of new normativity by inferring to (30) from (1a) and
(1b). We only get the same kind of normative claim that we presupposed in order to make sense
of (1b).
If this is correct, then the norm-relative view has a way to explain what is wrong with (1),
namely, that it begs the question against the skeptic by trying to prove something that it
presupposes. Because “rude” must presuppose a normative standard, (1) cannot show us how to
Section 4.2 Implications
What this means in the bigger scheme of things is that the norm-relative semantic view is,
all things considered, preferable to the pragmatic view. Not only does it do a better job explaining
the linguistic data, and our sense that thick concepts are a distinct category, but it does so while
also giving a straightforward, moral-skepticism-friendly, response to (1). Does this mean, then that
the semantic view has no real philosophical difference between the pragmatic view and the
norm-relative semantic view, at least in the grand scheme of things? I don’t think so. Here are some
reasons why.
First, even if thick concepts do not undermine the is/ought gap in a strong anti-skeptical
sense, there it has implications for the coherence and justification of our normative beliefs and
views. Specifically, suppose that thick concepts have the justification structure discussed above.
In other words, suppose that when I say that “S is courageous” I am evaluating her according to a
certain standard for virtuous behavior S, which is justified by a particular theory of virtue V. If I
believe that S is courageous, this means that I’ve implicitly or explicitly accepted these standards,
and this has implications regarding the coherence of my beliefs and commitments.
Suppose, for instance, that I find myself in a religious environment where terms like
“sinner,” “lustful,” “prideful,” “blasphemous,” “humble,” “pious,” “God-fearing,” “holy,” and
“obedient” are used to describe people and actions. If the norm-relative view of thick concepts is
correct, then when I sincerely say that some action is “lustful,” I am presupposing a standard for
virtuous character according to which sexual desire is negative. But it is possible that I could
routinely classify actions as lustful while at the same time denying this particular conception of
community, and yet I go on calling people prideful, blasphemous, or lustful, not recognizing that
I’m making evaluations when I do this. In a case like this, it would be easy for me to slide from
judging that S is prideful and therefore bad according to a standard that I reject, to judging that S
is therefore bad (in an all-things-considered sense). Compare this to my holding the belief that P
is FDA-approved, but also believing that the FDA is not trustworthy. If I held this belief, I would
need to be careful to qualify my beliefs regarding FDA approval, so as to avoid mistakenly
inferring that P is therefore justified or true.
The norm-relative view of thick concepts then, sheds light on views like that of Bernard
Williams (1985) and Elizabeth Anderson (1993), as well as other people interested in the
connection between thick concepts and our actual practices of evaluating. For instance, the
norm-relative view makes sense of the idea of an evaluative perspective, which can be understood as
something like a system of thick concepts that all derive from the same normative standard. Thus,
the thick concepts associated with Catholicism mentioned above might together constitute an
evaluative perspective, in the sense that they are all derived from the same conception of virtue. A
simpler example of an evaluative perspective might be the perspective from inside a soccer game.
Partly virtue of having the concepts “goal,” “foul,” or “out of bounds,” I know intuitively what is
good and bad in the context of soccer. In virtue of having the concept “goal,” I know that scoring
goals is something that I ought to try to do, and that doing so is good. I also know, in virtue of
having the concept “foul,” that I ought not to do certain things during the game.
Plausibly, then, thick concepts do seem to structure our evaluative perspectives by giving
us information about what is good or bad, right or wrong, in certain domains. This makes thick
“chocolate” or “athletic,” because only the former are part of an evaluative perspective in virtue
of their semantic content. Needless to say, the pragmatic view of thick concepts does not recognize
this aspect of thick concepts, and for this reason, we should consider the difference between the
norm-relative and the pragmatic views philosophically significant.
Section 4.3 The robust normativity objection
Suppose that the norm-relative semantic view is preferable to the pragmatic view along the
dimensions of explanation (it does the best with respect to the linguistic data) and simplicity (it
doesn’t require any problematic metaphysical leaps). This doesn’t mean that it’s necessarily the
best view of thick concepts, since it is still in competition with the default semantic view. Prima
facie, we have two clear reasons to prefer the norm-relative semantic view to the default semantic
view, both of which are the same reasons why we would initially prefer pragmatism to the default
view. Specifically, the default semantic view seems to commit us to the aforementioned
metaphysical leap from is to ought, and it has trouble explaining why the sentences discussed in
the defeasibility discussion earlier are not incoherent.
However, a defender of the semantic view might maintain that only she can explain
certain essential features of genuinely thick concepts such as their normative robustness, and the
fact that they seem to have correctness conditions that go beyond what we would expect of “merely
conventional” normativity. In this section, I’ll argue that the norm-relative semantic view can
explain how thick concepts are both normative in the conventional sense discussed, and higher
order correctness conditions. If this is correct, then we should also be able to conclude that thick
ought to qualify as robustly or genuinely normative. If all of this is correct, then we should prefer
the norm-relative semantic view to the default semantic view as well as to the pragmatic view.
Section 4.4 The objection from correctness conditions
The norm-relative semantic view is not new. However, it hasn’t received much defense,
and it is often dismissed on the grounds that relativizing the evaluative content of a thick concept
to something like the NORMS OF ETIQUETTE, or TRAFFIC LAWS, or the CONVENTIONAL
NOTION OF VIRTUE rules out the possibility of critique. Consider Hilary Putnam’s (2002)
objection to the norm-relative semantic view:
It is a mistake in its terms because as a “linguistic analysis” of the meaning of [thick] terms, it implies extremely implausible synonymy relations. For example, if
“courageous” were synonymous with, “fits in with normally accepted standards of courageous behavior,” then it would be a contradiction for someone (Socrates?) to say that the “normally accepted standards” confuse courage with rashness. Or to take Sen’s own example, if “barbarous” were synonymous with “has certain features usually associated with the notion of barbarity” then to show that capital punishment isn’t “barbarous,” it would suffice to point out that those features that capital punishment has in common with barbarity are not necessary and sufficient for barbarity as defined by
what is “usually associated” with it.14
If Putnam is correct, if we were to treat a concept like courage as though its normative content
were only conventional, then we would not be able to contest its meaning. For instance, the claim
that
(16) X is conventionally thought of as courageous, but X isn’t really courageous.
is contradictory if what it means for something to be courageous just is for it to be
conventionally thought of as courageous. According to Putnam, the only way to avoid this is for
us to assume that there is such a thing as an objective, or cognitive property of “being courageous,”
to which we can appeal in order to explain how the conventional view could be mistaken. But is
this a direct objection to the norm-relative view of thick concepts? Putnam here seems to assume
that if a thick concept is determined by convention, then the norm expressed by the concept isn’t
revisable, or isn’t subject to higher order correctness conditions (e.g., regarding evidence and
justification). But this assumes that being a convention and being revisable are mutually exclusive.
Call this assumption Conventional.
Section 4.5 Conventional
Let’s try to get a better sense of what Conventional says. The main idea seems to be that if
Conventional is true, then either (a) a concept’s meaning is conventional, or (b) the concept’s
meaning is revisable in light of a higher order standard, but not both. Thus, Conventional says that
the following cannot both be true.
(17) The meaning of courageous is whatever is conventionally accepted as courageous.
(18) The conventionally accepted conception of courage is mistaken.
The intuition is that if we stipulate that the meaning of COURAGEOUS is WHATEVER
IS CONVENTIONALLY ACCEPTED AS COURAGEOUS, then it doesn’t make sense to claim
that our conception of courageous is wrong, since whether it is wrong just depends on what
In order to get a sense of why this objection isn’t a problem for the norm-relative semantic
view, it would help to recall the discussion in 3.3. In that section, I began to make a distinction
between two sorts of “higher order critique.” First, there is the sort of critique that Vayrynen
mentions, where we might contest whether “lewd” has evaluation as part of its meaning. Second,
there is the sort of higher order critique where we critique the concept LEWD by way of critiquing
the normative standard to which it refers. An example of this sort of higher order critique was that
of criticizing the concept (or the standard expressed by the concept) FDA-APPROVED in light of
evidence undermining its claim regarding drug x. This second sort of critique might be best
described as a critique of the norm that a thick concept expresses.
Now, Putnam’s objection to the idea that COURAGE means “whatever is conventionally
thought of as courage” is ambiguous in light of this distinction. This is because he might mean two
things: first, he might mean that the concept courage means whatever people take it to mean, and
second, he might mean that the standard expressed by the concept COURAGE (e.g., that one ought
to express a certain amount of aggression in the face of danger, no more, no less). If we read the
account of COURAGE that he is critiquing in the second way, then it will be clearer why the worry
is misguided.
If Putnam means to criticize the idea that the norm expressed by the concept COURAGE
is determined by “whatever is conventionally thought of as courage,” then, because he thinks that
this option rules out Socrates’ critique of courage, he is committed to something like the following:
Conventional: A norm is conventional if and only if it is not revisable according to a higher order, non-conventional standard of correctness.
A norm that satisfies Conventional, it seems, will be one whose justification bottoms out—
(2011) distinguishes between conventional norms and other norms on a similar basis, proposing
that a norm is conventional just in case it is “grounded in social practices” in the sense that we
justify these norms by appeal to “the done thing.” Some of his examples include:
(19)
A: Why do you think that one must pass the port to the left? B: Passing the port to the left is just the way it's done around here.
(20)
A: Why do you think that one must not go naked on the beach? B: Going naked on the beach is just not done here.
(21)
A: Why do you say that one must black at funerals? B: Wearing black at funerals is just what we do here. (774)
Plausibly, these norms, passing the port to the left, and so on, are conventional norms in
Putnam’s sense. If Conventional is correct, then if we were to posit some sort of moral justification
for the social practice of passing port to the left, then it would cease to be a conventional norm.
Section 4.6 Against Conventional
But Putnam’s requirements for what makes a norm conventional are too strong, since many
of our conventions are subject to some higher order standard of justification or correctness without
thereby ceasing to be conventional. Or that is what I will try to argue in this section. First, let’s
consider some norms that appear, intuitively, to be both conventional and subject to higher order
The most obvious case here is legal norms. Take, for instance, the norm that one ought not,
legally, cross the street without using a crosswalk (i.e., to jaywalk). Intuitively, this norm is
conventional, in the sense that my decision not to jaywalk is something like this sort of action “is
not allowed here.” Further, whether this sort of crossing the street is permissible seems to be
entirely dependent on facts about local traffic laws, which are generally taken to be a paradigm
instance of this sort of convention.
Yet, it should also be clear that traffic laws are subject to higher order correctness or
justification conditions in the sense that we can revise our traffic laws if some turn out to be
problematic and need to be eliminated, or if we believe it would be better to incorporate new ones.
We can also revise the jaywalking law itself, perhaps by adding further specifications to which
kinds of street crossing constitute jaywalking (does this include bikes that cross the road? For
instance). It may even be the case that traffic laws are, at bottom, subject to moral justifications
(e.g., utilitarian justifications, for instance). But even if this is so, traffic laws, intuitively, will not
cease to be conventional.
We can apply this same schema to any conventional norm, including the norms of soccer,
baseball, or chess. Suppose that it’s a norm of soccer that throwing the ball constitutes a foul. This
is clearly a conventional norm, in the sense that it depends on either a formal or implicit set of
soccer rules. But soccer rules and the rules of baseball and chess are, as a matter of fact, also
subject to higher order justification or correctness conditions. We can coherently argue, for
instance, that the game of baseball would be better if one could get four strikes before going out.
Section 4.7 Implications of rejecting Conventional
Once we’ve rejected Conventional, we can return to the problem that we started with.
According to Putnam, the norm-relative view of thick concepts—a view which says that the
meaning of COURAGEOUS is WHATEVER IS CONVENTIONALLY ACCEPTED AS
COURAGEOUS —cannot accommodate our need to critique the concept of COURAGE on
grounds other than popular consensus. This, again, assumes that to define COURAGEOUS
according to convention is to thereby stipulate that it is not revisable. But this assumes too strong
a notion of “whatever is conventionally accepted,” since “whatever is conventionally accepted”
may be accepted on a rational basis, and may be subject to rational critique.
Recall the analogy with traffic laws. We can say that JAYWALK is defined according
to whatever is conventionally accepted as jaywalking, but it is implicit in this definition that by
“conventionally accepted” we mean “with reference to some conventional norm.” But so long as
we reject Conventional, this does not rule out that such norms will be subject to critique, again, on
rational grounds.
If this is correct, then the definition of COURAGEOUS as WHATEVER IS
CONVENTIONALLY ACCEPTED AS COURAGEOUS doesn’t rule out the option for critiquing
the conventional notion of courage. This is because, plausibly, in the same way that specifying that
JAYWALK is defined conventionally means that its content depends on facts about (conventional,
yet revisable) traffic norms, specifying that COURAGEOUS is defined conventionally means that
CHAPTER 5: CONCLUSION
The goal of this paper has been to defend the semantic view of thick concepts, and thereby
the idea that evaluative thickness is philosophically significant. In their haste to avoid certain
extravagant conclusions drawn from the semantic peculiarity of thick concepts, many prominent
philosophers have dismissed the category altogether, accusing defenders of thick concept theories
of mistaking ordinary social psychological facts about how we use certain terms with their
semantic content. But this dismissal goes too far, disregarding things that it shouldn’t.
My proposal in this paper has been that thick concepts as a class include any concepts
that, in virtue of their semantic content, rank something according to a specific normative standard,
according to certain descriptive features that it has. If I am right, then the category of thick concepts
includes many more than the paradigm virtue concepts like “courageous” and “cruel.” It also
includes concepts like “soccer foul” and “FDA-approved.”15 While many philosophers have
assumed that this view cannot accommodate our sense that we can be mistaken about the meaning
of a thick concept, I’ve tried to argue that this assumption is mistaken so long as laws, games, and
idiosyncratic codes of conduct are all, in principle, subject to higher order justification.
If this much is correct, then several things follow. First, we should reject the pragmatic
view of thick concepts, which either doesn’t accommodate the norm-relative semantic view, or
requires significant revision to do so. Second, we should reject the assumption that thick concept
arguments like the one discussed in the introduction undermine the is/ought gap in the way that
15 It does not include, however, concepts like “athletic” or “chocolate,” which distinguishes it from
has often been assumed. Finally, we should pay attention to thick concepts to the extent that they
are relevant for our knowledge practices. Specifically, there is a sense in which merely acquiring
a thick concept causes me to adopt an evaluative attitude toward something, even if it is
context-specific (e.g., soccer goals are good, jaywalking is bad). If this is the case, and if I am right in
thinking that all of the relevant normative standards are, in principle, subject to higher order
justification, then the norm-relative view of thick concepts is important in the sense that it gives
us a better framework for identifying, criticizing, and aiming to achieve coherence among our
REFERENCES
Anscombe, Elizabeth (1958) ‘Modern Moral Philosophy’ Philosophy Vol 33, No 124 (January), reprinted in Ethics, edited by Judith Thomson and Gerald Dworkin (Harper and Row, 1968).
Foot, Philippa. (1958) “Moral Arguments,” Mind 67, 502-13.
Lehrer, Adrienne (2009). Wine and Conversation. Oxford University Press.
McDowell, John (1985). “Values and Secondary Qualities.” In T. Honderich (ed.), Morality and Objectivity: A Tribute to J.L. Mackie. London: Routledge & Kegan Paul, 110–29.
Parfit, Derek. (2011). On What Matters: Volume One. Vol. 1. Oxford University Press, 2011.
Putnam, Hilary (1981). Reason, Truth and History. Vol. 3. Cambridge University Press, 1981.
Putnam, Hilary (2002). The Collapse of the Fact/Value Dichotomy and Other Essays. Harvard University Press.
Rawls, John (1955). “Two Concepts of Rules.” The Philosophical Review 64, no. 1: 3-32.
Searle, John R (1969). "How to derive ‘ought’ from ‘is’." In The is-ought question, pp. 120-134. Palgrave Macmillan, London.
Southwood, Nicholas (2011) “The Moral/Conventional Distinction.” Mind 120, no. 479: 761-802.
Väyrynen, P. (2013). The Lewd, the Rude and the Nasty: A Study of Thick Concepts in Ethics. New York: Oxford University Press.
